Reducing Inputting Errors in Profiles

Question

The error lists or Suggestions if you want to put a more positive spin on it seem to be getting longer as more and more errors are being discovered.  This lists are essential for correcting errors but it strikes me that there is very little being done at the data entry points to stop, or at least reduce, the number of data entries occurring. For some errors we do have the red banner that pops up when you try to save a profile which gives you the opportunity  to correct an error or ignore it if you know that the error is in another profile which you have yet to edit.

My first suggestion is that this red banner should be extended to checks on all input fields with the exception of the Biography field. This would be particularly helpful in the Gender field where a check could be made agains the entered first name but could equally applied to other fields particularly when a spouse's name is entered with the the prefix of either Mr or Mrs.

My second suggestion is that as the Suggestions lists are either ignored or they are not aware of their purpose by a lot of profile managers making a lot of unnecessary work for the Data Doctors. My suggestion is that when you open WikiTree a popup appears saying that you have x number of suggestions which require attention with a button to take you to your suggestion list.

There may vey well be other ideas for reducing data errors so I will throw this one open.

Comments

I'd like to see a small red clickable icon in the top bar if the system's got something to tell me, for example if my Suggestions report is not empty. That's what you see on more well-designed sites like Facebook or Goodreads. It's probably more efficient than a pop-up or a big red banner.

---

And you know what?  I tend to ignore little notification boxes - sometimes, because I can't be bothered, and sometimes because I just plain don't see them.  Now, a great big scarlet banner - you can't 'not see' that. ;)

---

There's a very big difference between a small notification and no notification at all, like what is currently the case with the Suggestions.

A small notification would have worked for me. I never look up my Suggestions unless I'm reminded of it, like I've been today with this thread. There were two errors there, but I have of course no idea if there have been others that's been fixed by Data Doctors..

Big red and yellow banners may work for a while, but tend to annoy users, and may eventually make them go away. I'm very bored with the dire warning about even touching a pre-1700 profile if I don't know what I'm doing. A Web interface littered with big ugly banners is counter-productive in the long run.

---

It appears that my example of using Gender as an example has caused a bit of an issue. The point I am trying to get across is that WikiTree needs better error checking at data entry and I thought that using gender and first names would be a good example on how to possibly go about it - apparently not.

As we are generally talking about gender now, non specific gender names are an issue and the suggestion list is full of entries where the gender is correct but the software thinks it’s an error due to the name so some form of verification at input would reduce the error count significantly. How you would do that effectively is s decision only the programmers can make or even if it is worth doing. Errors on entry are the responsibility of profile managers but they do need some help as we all make mistakes at some time.

Whilst it is easy to give reasons why it won’t work, which are often very valid, it would be good to keep positive in order to improve WikiTree whilst making more user friendly.

---

David, while I certainly can agree that WikiTree needs to be made more user friendly, I think that your name/gender example is a great way to explore how a good intention can lead to an adverse user experience. As a gender check on first name is bound to generate a large number of false alerts, users will become annoyed and desensitised to the message, and click it away automatically, even in the cases when there's a real reason for the warning.

Just as non-Americans, who come from cultures where multiple first names are often rather the norm than the exception, may get upset by getting warnings about "unusual data" every single time we enter a perfectly normal compound first name, and eventually don't even read the text of the warning, which actually may cover other issues.

These are classical "cry wolf" situations, which only marginally help improve the overall WikiTree data quality, at the risk of alienating users.

---

Gender was probably not the best example.   
But locations like someone trying to enter USA before 1776

Or Germany before 1871

Or a date that makes the parent under say 18

Or where you see the same parent as mother and as father

or where you see a male being listed as mother

There are a ton of issues that could be stopped at pont of entry and this is really needed with gedcom uploads.

---

I agree with the multiple first names. Virtually every entry I do has amultiple first name.  I have got into the habit of going Save Anyway without reading the message which might be for something else.

Answer 1

As a new WikiTree participant and recently Data Doctor, I accept the premise that members who are adding records are not aware there is a Suggestions tool to assist their entry. Nor are new members generally aware that the Research feature exist.

I expect more education on these tools and more focus on required field use and standard sub-headings and positions of sourced and non sourced data and reference use and format to use for different types of source reference records would assist the members to identify their own suggestions and with the recently created videos, correct many of  the basis errors.

I know all this stuff is out there; but, we must recognize that the majority of members are not seasoned genealogist and perhaps a "Read This First - Standards Doc" or "How to Best Standardize your WikiTree Profiles" could be sent out or pop-up with the new profile creation.  Too many links in this type of document cut down on what is actually read.

Ok, perhaps after I am seasoned more, I will hold a different position.

Comments

Dale, thank you for the notation on the Research feature.  I've been on Wikitree for three years (and a Data Doctor for a while) and didn't know about that.  It seems like it would make searching across sites easier with just one entry.  (Found it on my dropdown list a few places above my suggestions list).

Answer 2

David - add tag Improvements

Comments

Thanks Linda if have added the tag

Answer 3

Great ideas, David!  I don't know how much work that would make for Ales, but the red banner messages being expanded to include other errors makes a lot of sense to me.  

I would be for the suggestions list notice also, if there is away to turn it off.  Having it pop up initially would alert the manager that it exists. But there are some errors that will never be corrected such as an infant's gender that has died and no documentation is available to prove the sex.  It might get annoying when one has made all the corrections they can, but errors still exist on the list.  The only remedy would be to make them all false errors but I don't think that is necessarily the best approach, as someone might come along later with the evidence to correct an error.

Just my thoughts!

Comments

Carolyn, you can make the unknown infant gender go away by marking it a False Suggestion.

I do agree that there are some errors that will never go away.  People I've researched where a unique name can't be confirmed, or where a PM won't correct a LNAB problem (and I can't do that), etc.

Answer 4

To check gender against given name may initially sound like a great idea, but in practice it's futile. You have to build a huge list of given names, and all too soon you will run into exceptions. How do you suggest to handle a name like Jean, which is a male name in France, but a female name in Anglophone countries? Remember that WikiTree is pretending to be a global tree. I can give several examples of ambiguous given names in Norwegian, like Helge and Tore, which may be given to both boys and girls.

Programmers have an acronym for this kind of suggestions: TMWFTLB - Too Much Work For Too Little Benefit.

Comments

I check the suggestions report almost weekly for French epicene names. For instance, the system assumes Claude is male, though in France it is quite frequent for women also. So I check those and mark them false errors when appropriate, so that I'm sure the gender does not get changed by mistake. And yes, in French Jean is always male; Anne is male in Dutch (and occasionnally in French as well); pre-1700 you'll find girls called Philippe, Antoine, Eustache or Ambroise (all assumed male by the system); numerous boys have "Marie" among their given names (and, like Claude, the gender of "Marie Joseph" is impossible to guess without context).

---

We already have some criteria built because we have the suggestions of probably male, probably female, etc.  Couldn't that be used to make a yellow warning box (similar to the one for two first names)?

---

I do gender assessment by sampling wikitree database. You can lookup any name you want here.

http://wikitree.sdms.si/default.htm?report=ana1&FirstName=Fern&MaxNames=500

I do include all first names into assessment as you can see With Marie Joseph example.

http://wikitree.sdms.si/default.htm?report=ana1&FirstName=Marie+Joseph&MaxNames=500

I think it produces quite good gender assessment. By adding location to the mix, it could be improved or not, since we have a lot of migrations. Same goes for time component. Name usage changed over time.

But that would mean a lot of effort for minor improvement.

---

Aleš, don't get what those URLs are supposed to tell me. I only see a blank page, with a row of buttons on the left side.

---

Click the blue Analyse Names button on the left.

---

I agree, Aleš - I think your gender assessment by statistics works pretty well - in the circumstances - and might be more work than it is worth to tune any finer. But then we probably have less of the ambiguities in Sweden than in France.

I think the matching of name and gender is appropriate as a DB_error, where you can survey it as a separate task, and mark it as False Error when it is fals.

I don't think it is appropriate as a data check upon save - since there would be too many false warnings.

---

Good discussion. I agree with Leif!

Current automated banners already create a few false alarms that make editing clumsy. TMWFTLB - Too Much Work For Too Little Benefit examples include:

1 –Women died of childbirth complications more frequently in the past. You often have a child born in, for example,  1789 and the mother die in same year. A red warning banner shows every time you edit them & an extra step is needed (actually several as you need go back each time to confirm you have it right).  

2 – Males in my family fathered children out-of-wedlock & in-wedlock at roughly the same time. There are also a few twins. A red warning banner shows every time you edit them & an extra step is needed (actually several as you need go back each time to confirm you have it right).  

So please don’t add other checks that will require extra steps.

 

---

Just so you know, number 1 is corrected on Test server and will be added to production soon.

---

Really excellent news!  Looking forward to that change.

---

Jim Wiborg, I thought that was only males  in my tree, LoL.

Trish (adopted born Loretta Morrison

Answer 5

Never add the prefix "Mr." or "Mrs."  It adds no information to a profile, and makes a profile look like clutter.  A lot of errors, like the gender one, are old.

Comments

Never say never. These titles can indicate social status, such as in 19th century United Kingdom.

---

Social status is abhorrent to a USian whose ancestors left the UK in the 1600s--except the one who left in 1849.

---

True, but I don't think we should project our modern prejudices onto the profiles of people who lived in other times and places. :-)

Answer 6

There IS already a banner if you want to add a person without specifying the sex: "Attention, the person has no gender. Is that intended?" (a bit my wording, but the sense is given.

I agree to what someone said about names used in different countries for different sexes. Just two examples: Simone is in Germany a female name, but in Italy a male name. Nicola is in GB a female name, in Italy written with C a male name and in former Yugoslavia written with K a male name. But I don't even have to change the country to see a changing in the sex. In the 1800 there was a FEMALE name Christian in GB, and today it is a male name. And now?

If you want to diminish your suggestion report, simply click every week on it and see what you can do to reduce the number of suggestions. I don't think it is that good to put an alert on every single field while you create a profile. I'd get mad if I had to click away dozens of alert fields while creating profiles.

Answer 7

I think one warning banner with a list of what the possible issue is might work to begin with.  That would allow for faster implementation but may not tell the user the exact issue causing the problem.  That could come later.  I would hope most users would be able to figure it out from a prompt list.  (See my comment entered on the opening statement).

Comments

Alerts are fine as long as they're referring to real issues. They are mostly annoying and leading to warning fatigue when they raise large numbers of false alarms, like multiple given names, gender ambiguous given names, or whether you're qualified to edit pre-1700 profiles.

But please, give us more useful alerts, and be willing to tune or remove conditions which give rise to more than the occasional false alarm. Normal entry of data should never raise any concern. It may of course be debated what constitutes "normal" data entry. But given that WT is pretending to build a global tree, demand for some slack seems reasonable.

---

Warning banners that do not tell the user what the problem is, in an intelligible way, are immensely frustrating.

There are regularly postings in G2G about the warning banner for birth of a child (too long) after the death of a parent. It seems (from memory) that it is not always clear which parent and which child it refers to. It also pops up when it shouldn't because the algorithm isn't smart about dates that are marked as "before" or "after".

Oh, and also from what we regularly see in G2G: the alert that appears when you try to create a new profile with a non-existent date like April 31 does not seem to be very enlightening. People try over and over again - and then turn to G2G.

There is certainly scope for improvement.

Answer 8

Suggestion reports are created on my servers in Slovenia and WikiTree doesn't have access to it. It is also done using different database and programming language. So there is no way to simply add it into wikitree.

There were some ideas to integrate the suggestion reports into wikitree, but none were implemented until now. It is still completely separated from WikiTree. We might revisit this possibility.

Since Data Doctors project started, the awareness of errors in data increased and WikiTree added some data checking on saving a profile. Mostly date errors and some others. But this checking is done completely separated from WikiTree+. This checks reduced the inflow of new errors on new profiles, so it is a great thing.

But I don't think All checks will be done on save. OnSave checks can be improved and new ones added, but that should happen one by one with clearly defined algorithms and consensus in G2G. Having more warnings like multiple first names is not good, since it is country specific.

Comments

I understand that a seamless integration between those two systems are out of the question. But it might be possible for WikiTree to send a request to your server for the count of suggestions for each member that has made any contributions the last week. It's probably only a few thousand, cf. the "Club 100" list. Any member that comes up with a non-zero number of suggestions might get a notification with a link to the Suggestions. I think that would be an immensely useful improvement.

---

That's helpful to know and explains why the suggestion code couldn't be applied earlier in the process.  Thank you for the explanation.  Especially since the on save checks would need to be realtime and fast, such as where we started regarding names which would take too long to go from Wikitree to you then back.

Personally I look at my suggestion report several times a week and don't want any more email like that.  If anything like Lief is suggesting here is done, it also needs to look at who has opened their suggestions list and exclude them.  That would be worse than too many red or yellow boxes on the saving function.

---

When I wrote "notification" above, I was thinking more of a little red Facebook-style spot in the top bar with a hover-text like "You've got Suggestions", than an email. That would be quite sufficient for me as a reminder.

And please stop calling me Lief. My name is Leif.

---

Yes, like the little arrow now on the suggestions with a video.  That little red arrow would be fine with me - I agree it would help newbies be aware there is something to see (and fix) there.  Those who work their suggestions could ignore it.

---

Aleš, 

Is it possible to change the placement of the Suggestion Tab? Instead of being buried under My WikiTree Tab, could it be a main tab (or link, button whatever it is called) between My WikiTree and Profile Name-#, Call it "Research Needed" or "Check your Suggestions". Would this just be a matter of moving the link to the Individual Suggestion Report and not involve at lot of programming between the two systems?

I believe most WikiTreers don't realize this report is available for them and if it is easy to find, they may work on their suggestions.

Just a thought, sometimes the easiest solution is the most difficult to find.

---

WikiTree design is not under my control. Chris decided it will be there so it is there.

Answer 9

Perhaps to address outstanding issues we could have a template that Would generate a “research needed” item, and when present and containing the matching suggestion number would inhibit the alert.

So an alert would have 3 options: fix the issue, mark it false, or add the template. These last two could each have a button to get them done.

Answer 10

In my brief experience as a Data Doctor most of the errors I have encountered (regardless of type) involve GEDCOM imports.  I understand the process has been improved not to long ago so I may be encountering older profiles but it seems there is so much extra, unneeded stuff that is imported and dumped into a profile.

Comments

Um either too much or not a lot.  Not sure which is worse at least too much gives a starting point

---

Ditto on both your comments. I have done 2 challenge weekends now and I think these gedcom dump ins are the biggest offenders. So many with no source other than trees and they are not marked with a Unsoured tag even so tough to find , if it wasn't for all the other errors it creates!

Whilst fixing try to find at least one valid source to add.

Trish (adopted born Loretta Morrison

Answer 11

It would highlight to PM's if we add the link to their suggestions on any profile we fix

I leave a short message to say I visited the profile and why then I add the link into the message in the comment box. Last line

Appreciate any assitance you can give in fix these suggestions on your branch of the tree

All the best Janet ~ Data Doctor

Had a reasonable good response to leaving comments